24-hour AI news: ASML bets on the launch of Mistral and Wenxin X1.1, and the policy and implementation are issued simultaneously

In the past 24 hours (September 10), there has been dual progress in model and industrial policy in China, Baidu released Wenxin large model X1.1 and disclosed ecological data, and the Ministry of Industry and Information Technology proposed a special action direction of "artificial intelligence + manufacturing"; Abroad, ASML invested 1.3 billion euros to lead Mistral AI, and OpenAI landed a service partner in APAC and launched parental controls.

1. ASML leads Mistral AI, accelerating the formation of the European AI landscape

1. ASML announced an investment of 1.3 billion euros in Mistral AI as the core investor of its C-round financing of about 1.7 billion euros, holding about 11% of the shares and entering the strategy committee.
2. This round of financing has significantly increased Mistral's valuation, positioning it as a "European AI independent force", and forming in-depth synergy with semiconductor equipment leaders.
3. Industrial significance: The hardware-model-application chain is tighter, and Europe is striving for greater voice in large models and industrial AI.

2. Baidu released Wenxin large model X1.1 (deep thinking version)

1. Released at WAVE SUMMIT 2025 in Beijing on September 9, the official disclosure showed that compared with X1, the factual improvement was 34.8%, the instruction compliance was improved by 12.5%, and the agent ability was increased by 9.6%.
2. X1.1 has been fully opened on the Wen Xin Yiyan, Wen Xiaoyan and Baidu Intelligent Cloud Qianfan platforms, and the flying paddle core framework 3.2 has been launched simultaneously.
3. Ecological data: The number of Flying Paddle-Wenxin ecological developers has reached 23.33 million, serving 760,000 enterprises, indicating that the industrialization capacity of domestic large models continues to expand.

3. Ministry of Industry and Information Technology: Research and introduce a special action of "artificial intelligence + manufacturing"

1. According to the information at the press conference on September 9, a roadmap for the transformation of "artificial intelligence + manufacturing" will be formulated and a guide for AI application in the manufacturing industry will be released.
2. Focus on promoting key links such as high-end computing power chips, industrial multimodal algorithms, agents, humanoid robots and brain-computer interfaces.
3. Supporting measures include the construction of open source communities, the improvement of the standard system and the support of industrial funds, and the two-wheel drive of "intelligent industrialization" and "industrial intelligence".

4. OpenAI's layout and security governance in the Asia-Pacific

1. Service cooperation: Announced the first official service partner (Thinking Machines Data Science) in APAC, focusing on ChatGPT Enterprise training, customized applications, and landing operations.
2. Safety governance: According to media reports, new measures for parental control and protection of minors have been introduced to respond to compliance and social responsibility concerns.
3. Regional and compliance go hand in hand, reflecting the trade-off between security and usability in enterprise-level and public scenarios.

5. Domestic AI standards and industry activity dynamics

1. According to authoritative media reports, China has released a total of about 30 national standards for artificial intelligence, 10 new ones and 48 technical documents were established during the year, covering multimodality, agents, embodied intelligence and industry applications.
2. Implementation of "AI + energy": The National Development and Reform Commission and the National Energy Administration issued opinions to clarify the research direction of typical scenarios such as "AI + nuclear power/controllable nuclear fusion".
3. The Inclusion Bund Conference will be held from September 10th to 13th, releasing the "AI Ten Questions World", focusing on topics such as AGI, AI open source, AI medical care and robotics.

6. The online travel industry welcomes the transformation of "AI agents"

1. A number of overseas OTAs have cooperated with large model parties to explore "agent AI" from consulting to automated itinerary booking and customer service processing.
2. Intensified industry game: The platform emphasizes its own data and channel advantages, and hotels and consumers pay attention to the potential for disintermediation and cost reduction.
3. Short-term accuracy and risk control are still the thresholds, and in the medium term, it will affect the commission model and content distribution ecology.

7. Pharmaceutical R&D scenarios: large pharmaceutical companies open AI platforms

1. Multinational pharmaceutical companies launched AI drug discovery platforms for biotechnology companies, claiming to have accumulated more than $1 billion in models and data capabilities.
2. Platform-based openness helps small and medium-sized R&D entities reduce trial and error costs and shorten the iteration cycle.
3. Supporting mechanisms such as real-world data evaluation and regulatory communication to promote the translation efficiency of "from paper to clinic".

8. Alibaba Cloud's AI cloud share and industry collaboration (domestic market side)

1. According to the industry research summary, Alibaba Cloud ranks first in China's AI cloud market with a 1H25 share of about 35.8%.
2. Synergizing with the Qwen system and enterprise-level agent capabilities, MaaS and Agentization are regarded as the main line in the medium term.
3. The linkage between local life and cloud business is expected to strengthen the closed loop of "model-application-scenario" and improve business resilience.

9. Discussion of industry public opinion and evaluation methods: hallucination governance

1. Many overseas reports focus on the mention that "hallucinations are a systemic problem", focusing on evaluation incentives may lead to "blind guess is better than not answering".
2. Call for the introduction of indicators such as "uncertainty expression", "refusal incentives" and "source traceability" in the evaluation and training process.
3. Common enlightenment for model iteration at home and abroad: evaluation standards and training objectives need to be upgraded collaboratively.

10. Exhibitions and demonstration applications

1. Chongqing World Intelligent Industry Expo (September 5-9) focused on "artificial intelligence +" and intelligent networked vehicles, displaying more than 3,000 achievements.
2. The docking of industry, university and research has accelerated, and the number of landing cases of industrial agents, edge AI, and robot cluster scheduling has increased.

Frequently Asked Questions (Q&A)

Q: Why is it important for ASML to invest in Mistral AI?

A: The amount is 1.3 billion euros, the shareholding is about 11%, and it has entered the strategy committee. This is not only a financial investment, but also a vertical collaboration of "equipment-model-scenario", which marks Europe's active layout in the combination of model and industry.

Q: In what dimensions does Wenxin X1.1 improve compared with the previous generation?

A: The official gave three core indicators: factual improvement of 34.8%, instruction compliance improvement of 12.5%, and agent ability improvement of 9.6%. And it has been opened on the Wenxin Yiyan, Wen Xiaoyan and Baidu Intelligent Cloud Qianfan platforms.

Q: How will the "artificial intelligence + manufacturing" special action affect enterprises?

A: We will release a transformation roadmap and application guide, cooperate with open source communities, standards and financial tools, and give priority to accelerating the implementation of high-end equipment, industrial multimodality and intelligent agents, so as to reduce the threshold and cost of pilot-scale development.

Q: What trends do OpenAI's APAC service partners and parental controls represent?

A: The former emphasizes the landing of enterprises and talent empowerment (from "usable" to "usable"); The latter reflects the strengthening of platform responsibility and compliance, especially the refinement of minor protection and content governance.

Q: How will AI agents change the online travel ecosystem?

A: The evolution to end-to-end automation of "intention-plan-booking-after-sales" may compress the commission space of traditional intermediaries; In the short term, it is still limited by accuracy and risk control, and content and channel bargaining will be reconstructed.